Sei sulla pagina 1di 11

IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 26, NO.

4, JULY/AUGUST 2020 4400111

Wavelength Demultiplexer Designs Operating Over


Multiple Spatial Modes of a Rectangular Waveguide
Miri Blau and Dan M. Marom , Senior Member, IEEE

Abstract—Wavelength demultiplexers are instrumental in wave- Fig. 1-c) due to simplified wiring. Such an implementation
length division multiplexing (WDM) communications. In advance appears identical to its SMF counterpart yet contains mode di-
of space-division multiplexing (SDM) WDM communications, we
vision multiplexed (MDM) channels [4]. Current WDM mul-
investigate wavelength multiplexer/demultiplexer designs operat-
ing over multiplexed spatial modes simultaneously, when the modes tiplexers support the fundamental mode only, hence in order
are arranged along one transverse direction of a rectangular wave- to interface to an SDM transmitter, multiple, parallel WDM
guide. The demultiplexer waveguides’ modes match those of rect- multiplexers are required (matching the number of SDM chan-
angular core fibers, enabling simple and efficient edge-coupling. nels) as well as mode mux/demux elements [5]. With existing
We describe the challenge of demultiplexing mode division multi- single-mode wavelength multiplexers, SDM transceivers can
plexed (MDM) channels, and suggest guided optics solutions for
its implementation. We analyze the spatial–spectral interplay at interface to the common SDM-WDM fiber with either a single
output waveguide coupling due to mode dependent phase delays in multi-mode (or multi-core) output fiber containing the spatial
arrayed waveguide grating router demultiplexer design, and offer information (Fig. 1-b), which then needs to be mode demulti-
low mode dependent loss solutions. The proposed schemes allow for plexed, wavelength multiplexed and finally mode multiplexed,
efficient and compact MDM-enabled WDM multiplexing, paving or with multiple SMF output fibers, to be wavelength multi-
the way to SDM-WDM communication networks.
plexed and mode-multiplexed together (Fig. 1-a). Both schemes
Index Terms—Fiber optics and optical communications, planar reliant on multiple wavelength multiplexers are cumbersome.
lightwave circuits, multiplexing. Research leading to a simple wavelength multiplexer operating
I. INTRODUCTION over MDM has been relatively scarce. This fact can be explained
by two main challenges: (i) coupling spatially intricate, circu-
NTRODUCING space-division multiplexing (SDM), on
I top of the established solution of wavelength division
multiplexing (WDM), attracts the attention of the optical fiber
larly symmetry fields to a planar waveguide platform and (ii) the
interplay between spatial mode structure and spatial dispersion
when attempting to couple to an output demultiplexed wave-
communication research community as a capacity multiplier guide (resulting in irregular passband shapes, modal crosstalk,
that addresses the looming capacity crunch in today’s single and mode dependent losses-MDL). Consider the latter point,
mode fiber (SMF) based networks. SDM-WDM networks and when coupling spatially dispersed multi-mode light to an accept-
their realization [1], [2] are expected to economically scale the ing stationary waveguide. Let the dispersed light be defined by
transmission capacity in support of the expected traffic growth  
ahead, after all other capacity-enhancing measures have been dx
ψl x + (ω − ωc ) , y (1)
exhausted. SDM transceivers, preferably on a single photonic dω
integrated circuit (PIC), spanning spatial superchannels will fur-
ther drive costs down via resource sharing of a single laser source where ψl (•) is the spatial mode profile of the l’th mode and
and reduce receiver DSP load through common impairment its argument defines the mode position as a function of carrier
estimates [3]. It is further desirable that the SDM transceiver frequency ω (with ω c being the center carrier frequency).
interface to a single SDM fiber (i.e., multi-core, multi-mode, Note that using this ideal spatial dispersion definition, without
or their combination), and be subsequently wavelength multi- any modal dependence, implies all the modes are identically
plexed onto the SDM-WDM transmission link (as depicted in dispersed in the x-direction. The accepting waveguide supports
the same spatial modes yet stationary in space, ψl (x, y). Hence
Manuscript received November 28, 2018; revised February 27, 2019; ac- the power coupling of the dispersed light to the accepting
cepted March 25, 2019. Date of publication May 22, 2019; date of current waveguide is defined by the overlap integral:
version July 15, 2019. This work was supported in part by the European Com-
   2
munity’s Seventh Framework Program (FP7/2007-2013) under grant agreement  dx 
619732 (INSPACE), and in part by the ISRAEL SCIENCE FOUNDATION 2 
|ηk l (ω)| = ∫ ∫ ψk x + (ω − ωc ) , y ψl (x, y) dxdy 

under Grant 473/17. The work of M. Blau was supported by Shulamit Aloni dω
Fellowship provided by the Ministry of Science, Technology & Space, Israel. (2)
(Corresponding author: Miri Blau.)
The authors are with the Applied Physics Department, The Hebrew Univer-
sity of Jerusalem, Jerusalem 91904, Israel (e-mail:, miri.blau@mail.huji.ac.il; The coupling should be calculated for each frequency, due
danmarom@mail.huji.ac.il). to spatial shifts of the dispersed modal positions, and between
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. all combinations of input and output modes (hence the k, l
a
Digital Object Identifier 10.1109/JSTQE.2019.2909166 subscripts, denoting mode pairs). Coupling of LP01 , LP11 , and

1077-260X © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
4400111 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 26, NO. 4, JULY/AUGUST 2020

Fig. 1. Interfacing options from SDM transceivers to a WDM-SDM common communication fiber. (a) Single mode interfaces, requiring multiple external
single-mode wavelength multiplexers (matching the number of modes, M), followed by mode multiplexing. This arrangement presents a fiber management
challenge. (b) Multi-mode interfaces, requiring spatial demultiplexing feeding multiple single-mode wavelength multiplexers and finally mode multiplexing. Both
arrangements (a) and (b) offer opportunities for component integration. (c) Multi-mode interfaces followed by wavelength multiplexing operating over multi-mode
transmission. Fiber connectivity and component count is as simple as today’s single mode counterparts.

the LP01 and LP11 mode field distributions, this design cannot be
scaled to support higher order modes (either symmetric or asym-
metric). In [10], spatially demultiplexed modes were fed into a
single multiple-input/multiple-output AWGR demultiplexer fol-
lowed by mode multiplexing to achieve WDM demultiplexing
operating over SDM. Each SDM channel feeds one of the in-
puts of the AWGR, instead of a spectral channel. Hence this
method is limited to low mode and wavelength channel counts.
Finally, in [11], a multi-mode AWGR design was presented and
its performance analyzed. The fundamental problem of modal
dispersion in AWGR was presented, and for this reason the
fabricated MM-AWGR realized low spatial dispersion and very
coarse WDM channel spacing of 50 nm (at center wavelength of
850 nm).
To address the challenges of wavelength demultiplexing mode
multiplexed channels, we consider spatial modes defined by
Fig. 2. Power coupling between dispersed modes and stationary modes, sim-
ulating wavelength demultiplexing for MDM channels. rectangular waveguides constrained to be single mode in one
direction (width, defined by lithography) and multi-moded in
height (Fig. 3). This waveguide configuration can be designed
b
LP11 modes as a function of spatial (hence frequency) dis- to match the mode structure of rectangular core fibers (e.g.,
placement result in ill-shaped passbands and modal crosstalk core dimensions of 6 × 32 μm, for index contrast of 0.3%)
(Fig. 2). These occur due to the varying overlap between spa- [12], and allows for direct interface to this SDM fiber vari-
tially displaced (i.e., spectrally dispersed) and stationary wave- ant and realization of an MDM demultiplexer. By dispersing
guide modes, being strongly dependent on the specific field the signals along the single mode axis, the coupling to the de-
distribution of the spatial modes, and crosstalk arises since or- multiplexed output waveguides is devoid of spatial structure
thogonality is not preserved for the offseted fields, in this case which is the source for the ill-shaped passbands and modal
b
LP01 and LP11 . crosstalk. Our design goal is to have equal spatial dispersion
Nevertheless, there have been some strides in creating de- over all guided modes, allowing for wavelength demultiplexing
vices supporting wavelength multiplexing for MDM. In [6]–[8], MDM channels without introducing excessive mode dependent
a mode demultiplexer followed by multiple wavelength demul- shifts and losses. Equal spatial dispersion can be obtained in a
tiplexers were demonstrated on a single silicon PIC, for SDM free-space arrangement, using lenses and a diffracting grating.
and WDM co-demultiplexing. This functionality is depicted in Planar guiding devices, as AWGR and Echelle grating (Fig. 3-
Fig. 1-a. However, this is not the functionality we desire, our center/right) face the challenge of each propagating mode ac-
goal is to wavelength demultiplex MDM signal using a single quiring different phases due to their corresponding effective
WDM demultiplexer, without the need to mode demultiplex the indices. The different mode propagation constants within the
MDM signal, as shown in Fig. 1-c, avoiding the use of multiple planar device give rise to modal dependency in their spatial dis-
demultiplexers. Ref [9] presented a single arrayed waveguide persions. This undesired modal dependency must be minimized
grating router (AWGR) which, again, combined mode and fre- to reduce mode dependent wavelength shifts in the demulti-
quency co-demultiplexing. The AWGR supported the funda- plexer performance.
mental LP01 mode as well as only one of the degenerate LP11 Both Echelle grating and AWGR-based demultiplexers can
modes (oriented in the dispersion axis). However, since the mode disperse the propagating modes by imparting wavelength
multiplexing in this AWGR is based on the space symmetry of dependent delays across the beam aperture. It is the additional
BLAU AND MAROM: WAVELENGTH DEMULTIPLEXER DESIGNS OPERATING OVER MULTIPLE SPATIAL MODES 4400111

Fig. 3. Rectangular core fiber and demultiplexers. Left: RCF and its supported, polarization degenerate spatial modes. Center: AWGR-based demultiplexer,
designed to disperse the modes along the single mode direction. Right: Echelle-based demultiplexer.

mode dependence in these delays that impacts the demulti- free propagation region, and at the output surface excites an
plexer performance. As guided modes’ effective indices always array of waveguides, each longer than the other by ΔL. The
span the range between the core and cladding refractive in- waveguide array routes the identical copies of the input sig-
dices (ncladdin g < nef f < ncor e ), the Echelle modes would nal (containing all wavelength channels now in a single mode
exhibit greater spread as they are confined in one direction waveguide array) to a second free propagation region. Each
only within the free propagation region (n2ef f = n2cor e − κ2y d2 , waveguide in the array acquires a unique phase delay, differ-
with multiple spatial modes wavevector’s projection in the y ing from one waveguide to the next by a fixed phase difference
direction, κy , are supported, while d is the guiding layer’s Δϕ = 2πnef f ΔL/λc . Here λc is the center design wavelength,
height), whereas the effective indices in two-dimensional wave- and nef f is the effective index of the waveguide. The length
guide confinement of the AWGR are more closely packed difference ΔL is chosen such that the phase delay is equal to
(n2ef f = n2cor e − κ2y d2 − κ2x a2 , with multiple spatial modes and 2π·m for the center wavelength, where the integer m is the
κy supported and a single wavevector projection in the x direc- diffraction order. The free spectral range (FSR) is defined as the
tion, κx , and a is the waveguide width. Since we wish to reduce frequency period of the AWGR, and is inversely proportional to
the propagation delay spread in the demultiplexer designs, we the waveguide length difference and hence the diffraction order:
will pursue AWGR-based demultiplexer designs in the remain- F SR = νc · nef f /(m · ng ), where νc is the center carrier fre-
der of the paper (on the most part). quency in Hertz and ng is the group refractive index. The FSR
In this paper we propose wavelength demultiplexer designs must be larger than the frequency band spanned by all channels
for MDM channels, whose modal information is perpendicular to enable individual wavelength access.
to the axis of dispersion. We outline the performance bounds The light emerging from the waveguide array into the sec-
caused by the addition of multiple spatial modes. We suggest ond star coupler constructively interferes, i.e., is focused, at a
realizations for a wavelength demultiplexer operating over mul- particular position x depending on the relative phase delays ac-
tiplexed spatial modes by minimizing the modal dependency of quired in each waveguide of the array for respective wavelength.
the spatial dispersions. Finally, we show numerical test cases The lateral focus position shift obtained, or spatial dispersion,
for the suggested realizations, and discuss the advantages and is the key to demultiplexing and the output waveguides will be
disadvantages of each solution. For the first time, to the best located to collect specific wavelength channels. The impinging
of our knowledge, a viable wavelength demultiplexer for MDM light is collected by output waveguides, which completes the
is proposed for communication wavelength band and rate. The demultiplexing operation.
proposed solutions are scalable to higher rates, as well as higher In the case of waveguides that are multi-moded, we express
mode counts, enabling efficient WDM multiplexing over spa- the acquired phase delays per propagating mode and frequency.
tially multiplexed channels in a passive device. Assuming the AWGR is designed for the fundamental mode and
center frequency, we derive the acquired phase delay for every
II. MULTI-MODE AWGR DEMULTIPLEXER UTILIZING frequency component and mode l (see Appendix A):
RECTANGULAR WAVEGUIDES   
dβ (ω, l) 
The operating principle of the multi-mode AWGR demulti- Δϕ (ω, l) = β (ωc , l) + (ω − ωc ) ΔL
plexer is identical to that of conventional single-mode AWGR dω ω c
demultiplexer [13]–[15]. We first describe the operating princi-  (l) 
(l)
ple of conventional AWGR demultiplexer having single mode nef f ng (ω − ωc )
= 2πm (0)
+ (0) , l = 0, 1, . . . , M − 1
(SM) waveguides based on the analysis of [16], and defer the n n ωc
ef f ef f
full analysis of the multimode AWGR to Appendix A. (3)
Let an input single mode waveguide carrying wavelength
multiplexed channels feed a first free propagation region hav- where β and dβ/dω are the propagation constant and its first
ing confocal input-output surfaces. The guided light from the derivative and M is the mode count. The acquired phase delay
input waveguide now freely diffracts/diverges within the first depends on each spatial mode’s effective index and group in-
4400111 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 26, NO. 4, JULY/AUGUST 2020

Fig. 4. Left: AWGR’s input free propagation region, with significant physical parameters highlighted. Right: Depiction of output fields at the second free
propagation region, having different wavelengths, modes, and diffraction orders. Gray rectangles denote output waveguide locations for collection of demultiplexed
channels.

dex (see eq. 3). As an outcome, the spatial modes are spatially hence the difference:
displaced (i.e., dispersed) on the output plane as a function  (0) (l)

of both frequency and spatial mode. For wavelength demulti- nef f − nef f
x (ωc, 0) − x (ωc, l) = xF S R j−k+m (5)
plexing operation the spatial mode dependence is undesired, as (0)
nef f
it introduces mode dependent wavelength shifts to the output
waveguide, similar to polarization dependent wavelength shift From the equation above we can analyze the two param-
resulting from birefringence of SM AWGR devices [17]. The eters influencing the modal shift: The effective index spread
extent of the spatial mode wavelength shift will determine the and the diffraction order. For the conventional assumption
AWGR demultiplexer performance. of identical diffraction orders, k = j = −m, the modal off-
The spatial-spectral position of each mode at the AWGR set at the AWGR output plane is: x(ωc, 0) − x(ωc, l) = m ·
output plane focuses according to (see Eq. A12 in Appendix): (0) (l) (0)
xF S R (nef f − nef f )/nef f .
Ψl (x, ω) ∝ A depiction of the dispersed fields at the output of the second
   (l)
 free propagation region is shown in Fig. 4-right, assuming equal
(l)
1 λ0 R nef f ng (ω − ωc ) ng for all modes. Wavelengths are linearly displaced along the
G x− j + m (0) + m (0)
Δ nslab · p n n ωc spatial coordinate, realizing wavelength demultiplexer. Since
ef f ef f
(4) each spatial mode has a slightly different effective refractive
index, an additional offset will be present per mode. For demul-
Here Ψl denotes the field of mode l, whose position depends tiplexing action, we desire having many isolated wavelength
on the diffraction order, j (where j = −m is the designed grat- channels within the FSR.
ing diffraction order), the free propagation region size, R, the For the multi-mode AWGR to function as a wavelength de-
(l)
array pitch, p, the waveguide effective index nef f and the aver- multiplexer, all the modes pertaining to a particular wavelength
age effective index of the free propagation regime nslab .1 The need to couple into a common stationary multimode waveguide
optical carrier frequency is ω, and G(•) is a Gaussian func- (yet SM in the dispersion axis). Since the modes are spatially
tion approximating the waveguide’s fundamental mode in the dispersed, they will each have unique coupling efficiency to the
dispersion axis, and we find the mode position from its argu- common waveguide due to the spatial offsets, leading to MDL.
ment. We see that the position at the output becomes both mode Nevertheles there is no modal crosstalk within our wavelength
dependent and frequency dependent (the required spatial dis- demultiplexer, since the dispersion axis is orthogonal to the one-
persion) and excites multiple diffraction orders. For example, dimensional spatial content.2 We utilize the restrictive MDL
the fundamental mode l = 0 at the center frequency ω = ωc definition, i.e., the ratio between maximum and minimum loss
(0) (0)
will be located at x(ωc , 0) = xF S R (j + m · nef f /nef f ) where modes [18], although the standard deviation definition will give
xF S R = λ0 R/nslab · p is the displacement between the AWGR relaxed results. The standard deviation MDL metric is more ap-
diffraction orders, which yields x(ω c ,0)=0 for j = −m diffrac- propriate to device cascades. We wish to design the multi-mode
tion order, while mode l at the same frequency will focus at dis- AWGR such that the spatial modal spread, x(0)−x(M-1) (where
(l) (0)
placed position x(ωc , l) = xF S R (k + m · nef f /nef f ), where M-1 is the highest order mode supported), limits the MDL to
we substitute k for j in the diffraction order. For a wavelength an acceptable value. We choose 1 dB as our MDL target, place
demultiplexer we wish to minimize the modal positional shift, the common waveguide in the center, and derive a bound on the

1 Note that the different spatial modes experience slightly different, mode
dependent effective refractive indices with the free-space regime. This very 2 Note that there may be mode mixing at the fiber/planar device interface due
slightly effects the diffraction rate, and hence the number of excited waveguides, to spatial mode mismatch. Unraveling such mode mixing of the device interface,
as well as the absolute phase delay acquired by each mode, which bears no together with that experienced in fiber microbends, is possible at the receiver
impact. with MIMO-DSP.
BLAU AND MAROM: WAVELENGTH DEMULTIPLEXER DESIGNS OPERATING OVER MULTIPLE SPATIAL MODES 4400111

modal offset (as a function of the Gaussian mode size, Δ): III. SOLUTIONS FOR DECREASING WDM CHANNEL SPACING
  IN MULTI-MODE AWGR DEMULTIPLEXER
2 ((x (0) − x (M − 1)) /2)2
|ηcou plin g | = exp − We focus our design effort on the conventional telecom band,
Δ2
1528–1563 nm spanning ∼35 nm. The C-band width sets the
≥ −1 dB ⇒ x (0) − x (M − 1) FSR and maximal diffraction order at m ≤ 45. We target a
 200 GHz (Δλ = 1.6 nm) channel separation. Substituting the
≤ 2Δ 0.1 ln(10) ≈ 0.96Δ (6) spectral parameters (central wavelength and spectral separa-
using the same diffraction order (k = j = −m in Eq. 5) we find tion) in Eq. 10 we obtain the minimal value for the product
the condition: m · N ∼ 4250.
 Hence, our condition for spatial modes effective index spread
(0) (0) (M −1)
m · N ≤ 0.96 · nef f / nef f − nef f (7) (0)
(derived from Eq. 7) is nef f − nef f
(M −1)
≈ 2 × 10−4 , a very
which ties the spatial offset to the fundamental parameters of tight effective index range. In this section we suggest five ap-
the AWGR. N is defined as the effective number of arrayed proaches for minimizing the modal spread and obtaining an
waveguides (see Fig. 4-left): N = λ0 R/(nslab Δ · p), i.e., far AWGR supporting multimode channels. We describe each ap-
field mode size divided by pitch of the output waveguides. In proach and discuss the pros and cons of each.
deriving Eq. 7 we ignored the effect of group velocity mismatch
between modes, hence there is a spectral dependency, and we A. Very Low Index Contrast Waveguides
will need to define the minimal acceptable bandwidth for the Decreasing the modal spread can be obtained by decreasing
MDL limit we established. the core/cladding refractive index contrast to a level where the
In a similar way, we need to establish the acceptable minimal modal spread is small enough to enable sufficient spectral sep-
crosstalk to the adjacent wavelength channel (we set at −30 dB), aration as well as low MDL. The drawback of this approach is
whose waveguide is at position xch = (dx/dλ) · Δλ, where that low index contrast will result in large waveguide dimen-
(l) (0)
dx/dλ = ng mR/(nslab p · nef f ) is the mode-dependent spa- sions, providing poor overlap with rectangular core fiber, and
tial dispersion, and Δλ is the demultiplexer channel separation, necessitate very large bend radii for keeping bend losses low
which we wish to minimize. The adjacent wavelength wave- (leading to increased chip size).
guide position should therefore satisfy:
  B. Selective Mode Excitation in a Highly MM AWGR
2 xch 2
|ηcr osstalk | = exp − 2 ≤ −30 dB
Δ Alternative path to decrease the effective index span is to de-
 sign the waveguides of the AWGR to support a larger number
⇒ xch ≥ 6 ln(10)Δ ≈ 6.9 · Δ (8)
of modes and excite only a subset (the first M) of spatial modes.
This waveguide position is also limited by the modal spread; This reduces the spread of effective refractive indices especially
To avoid crosstalk the distance between channels must be mea- for well confined modes whose refractive index monotonically
sured from the highest mode of one spectral frequency to the converges to ncor e . In a rigid planar waveguide geometry, ran-
fundamental mode of the next channel. Hence the effective min- dom mode coupling is less likely. However, scattering from
imal offset xch is increased by x(0)-x(M-1). By equating the sidewall roughness may induce energy transfer to unused modes,
positional shift due to wavelength spacing of the AWGR de- add mode dependent loss, and mode mixing.
multiplexer and adding the modal spread (Eq. 6), we can obtain
the condition for the channel separation: C. Weakly Coupled Multi Core AWGR
xch ≥ 6.9 · Δ + (x (0) − x (M − 1)) ∼
= 7.8 · Δ (9) Another option is to use weakly coupled multi core waveg-
uides (vertically stacked) for realizing the AWGR. The coupled
From which we obtain:
waveguides support supermodes, whose structure resembles the
m · N ≥ 7.8 · λc /Δλ (10) rectangular core modes. The effective refractive indices of the
supermodes will be relatively closely grouped and become more
Thus, we have two conditions relating to the product m·N.
tightly bound as the separation increases. Of course, the limit
One condition tries to minimize it (the modal spread, Eq. 7) and
of this approach is simply a vertical stacking of single mode
one tries to maximize it (for denser wavelength spacing, Eq. 10).
AWGR demultiplexers, which is effectively another form of
The significance of this term relates to the resolving power of the
spatial diversity. This method requires very precise deep and
AWGR, which is determined by the total path length difference
vertical etching of the waveguides during fabrication.
of the waveguides, which is easily shown to relate to m·N. Let
us try to estimate some reasonable values for the AWGR design:
(M −1) D. Modal Offset Compensation Through Mid-Way Mode

10−3 , we
(0) (0)
let λc = 1550 nm, nef f
1.5, and nef f − nef f
can equate the values of m·N from the two conditions above Conversion
(Eqs. 7 and 10) and obtain Δλ∼6.4 nm∼800 GHz, which is too As the modal phase offset is accumulated when propagating
coarse for practical use. Hence it is paramount to decrease the through the device, we can compensate for the modal delay vari-
mode spread. ation by converting the highest order mode to the fundamental
4400111 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 26, NO. 4, JULY/AUGUST 2020

(and vice versa) half-way through the AWGR. This way the Δnef f ≈ 3.3 · 10−2 . As discussed above, larger modal sepa-
modal phase delays acquired will be equalized. This conversion rations are easier to obtain in one dimensionally constrained
technique is realized successfully for polarization offset in sin- waveguides, i.e., slab structures, lending this technique more
gle mode AWGR due to birefringence with a half wave plate amenable for an Echelle grating solution. Furthermore, it is pos-
[17]. However, such specific mode conversion is more difficult sible to obtain equally spaced refractive indices by designing the
to create for spatial modes, and will be needed for multiple pairs guiding refractive index profile to be parabolically shaped [20],
simultaneously with high mode counts. Mode converters utilize [21] preferably in high index materials e.g., SiON or highly
long period grating (LPG) for phase matching the effective in- doped glass [22], [23].
dex difference [19], which requires long interaction length as
well as unique LPG for each effective index difference or spatial IV. NUMERICAL EVALUATION OF TEST CASES
mode pair.
In this section we demonstrate by numerical simulations po-
tential solutions of the five options for SDM-supported wave-
E. Modal Offset Compensation Through Multiple Diffraction
guide grating routers described previously. All demultiplexer
Orders
cases are designed to support 4 spatial modes with 1.6 nm spec-
Since the AWGR radiates with multiple diffraction orders, tral separation in the C-band (200 GHz spacing). The first chan-
we can take advantage of them and compensate the modal offset nel is at 191.6 THz, the last channel is 195.8 THz with 22
using the neighboring diffraction orders. Instead of all modes channels supported. We set the channel bandwidth for the MDL
utilizing the same diffraction order and attempting to minimize limit to be 37.5 GHz.
the modal spread with respect to the spatial dispersion, here we Finite element method was used to numerically solve the
increase the modal spread with respect to the spatial dispersion effective and group mode indices of the spatial modes in the
and allocate a diffraction order per spatial mode such that they waveguides, while the output of the multimode AWGR is ob-
overlap in space. To obtain this compensation, the modal offset tained analytically using Eq. 4. The coupling to the output wave-
has to match the diffraction order spacing, xF S R , such that guide is then evaluated by performing the overlap integral per
successive orders overlap. Using the modal offset expression at mode and frequency. In each numerical case, we plot the channel
the AWGR output (Eq. 5), and setting it zero, we have: characteristics of the two adjacent wavelength channels located
 (0) (l)
 at the center of the C-band.
nef f − nef f
x (ωc, 0) − x (ωc, l) = 0 = xF S R j − k + m (0)
nef f A. Very Low Index Contrast
(11) For a low refractive index difference waveguide we as-
As the diffraction orders j and k are integers (with j as- sume a silica-based AWGR, with very low core doping. For
signed to the fundamental mode and equaling -m), we require nclad = 1.44398 with Δin dex ∼ 0.01%, height = 152 μm and
that the last term in the parenthesis be an integer as well, for width = 14 μm, the waveguide supports four spatial modes
any spatial mode l. This condition can only be satisfied if the in the vertical direction and is single moded in the horizontal
modes’ effective indices are equally spaced. We define the ef- direction (as required), see Fig. 5. The computed effective index
spread is nef f − nef f ≈ 2 × 10−4 . The output of a multimode
(0) (3)
fective index difference between two adjacent modes as Δnef f ,
(0) (1) AWGR based on this waveguide is also presented in Fig. 5.
i.e., Δnef f = nef f − nef f , hence the numerator evaluates to
(0) (l)
Although the modal spread and the spectral separation are ac-
nef f − nef f = l · Δnef f . The integer criterion for the last term ceptable, the dimensions of the waveguide will make coupling
in the parenthesis of Eq. 11 can be refined, requiring it to match in and coupling out a difficult and lossy process, in addition to
the mode number l, from which we find: large chip size and large bend radii due to the large mode size
(0)
nef f − nef f
(l) of the supported modes of the waveguide.
l · Δnef f
m (0)
=m (0)
nef f nef f B. Selective Mode Excitation in a Highly MM AWGR
(0) Next, we examine a higher refractive index difference wave-
nef f
= lΔnef f = (12) guide, supporting a high mode count in a small sized waveguide
m core. We considered very high index glass core [22], [23], with
which further means that the utilized diffraction orders are at ncore = 1.78, nclad = 1.444, width = 1.8 μm and height = 25 μm,
k = −m + l, such that the fundamental mode is at order −m, the waveguide supports more than 15 spatial modes in the ver-
the second mode at −m + 1, etc., yet they overlap in space at tical direction. The refractive index range of the first 4 modes
x = 0 for the center frequency ωc . However, since the spatial is within 2 × 10−4 , according to the limit we set in Section III.
dispersion term dx/dλ further depends on the unique group index The output of the AWGR is shown in Fig. 6. The drawback of
of each mode and the utilized diffraction order, the compensation this approach is the possible crosstalk to unused modes, which
at frequencies away from the center is reduced. will cause loss that can be strongly mode dependent, as well
The use of multiple diffraction orders and the design require- as mode mixing between the subset of excited modes (which
ments for their overlapping in space are thus equally-spaced can be compensated by MIMO processing at the receiver). As
refractive indices and relatively large modal spacing. We can the waveguide is designed to be single moded in the bending
(0)
estimate the latter by assuming nef f
1.5 and m = 45, hence direction, the only source of crosstalk will be roughness of the
BLAU AND MAROM: WAVELENGTH DEMULTIPLEXER DESIGNS OPERATING OVER MULTIPLE SPATIAL MODES 4400111

Fig. 5. Right: 4 modes of a low index contrast waveguide for SDM enabled WGR. Middle: WGR output simulation. Left: two neighboring spectral channels at
the WGR’s output

Fig. 6. Right: A single spectral channel at the WGR’s output. Middle: subset of 4 first modes of a high index contrast waveguide. The waveguide supports >15
modes. Left: WGR output simulation. Left: AWG transmission. Spectral channel crosstalk <30 dB and MDL <1 dB

Fig. 7. Right: A single spectral channel at the WGR’s output. Middle: 4 supermodes of the coupled waveguides. Left: WGR output simulation.

waveguide walls and imperfections in the multimode direction. D. Modal Offset Compensation Through Mode Conversion
This crosstalk should be relatively weak.
Our two last suggestions are designed to compensate the
modal offset, thus allowing wider refractive index spread.
C. Weakly-Coupled Multi-Core AWGR First we suggest to convert the fundamental mode to the high-
The third case is of weakly-coupled multi-core waveguides. est order mode halfway through the waveguide array. This way
This solution requires precise fabrication in which multiple AW- the modal spread will be compensated, as the accumulated phase
GRs will be inscribed on top of each other. The distance between of the fundamental and highest order modes will be equalized
the cores controls the effective super mode index, which al- at the end of the array. However, this solution adds complexity
lows us to engineer the coupling between the cores together and potential loss due to the addition of the mode converters on
with the effective index spread. Assuming nclad = 1.44398, each waveguide of the array, and is not easily scaled to higher
Δin dex ∼ 0.3%, height = width = 5 μm and 18 μm pitch be- mode count.
tween the cores, the effective index spread is 2 × 10−4 , see This solution will not compensate the modal offset com-
Fig. 7. However, this solution is complicated to fabricate and pletely, when only the extreme modes are swapped. Yet it
not trivial to couple to and from without adding to IL and MDL. will decrease the modal spread which is determined by the ex-
4400111 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 26, NO. 4, JULY/AUGUST 2020

Fig. 8. Right: 4 modes of a multi-moded waveguide in which the modal spread is compensated by mode converters. Left: WGR output simulation. Note that
first and fourth modes are perfectly compensated, and hence overlap and are located at the channel center.

Fig. 9. Right: 4 modes of a multi-moded waveguide in which the modal spread is compensated through multiple diffraction orders. Left: WGR output simulation.

treme modes. Full compensation will require additional mode effective index spread (see Eq. 11,12). Designing the demulti-
converters, whose number will scale as half the number of plexer in line with this scheme, results in full compensation of
propagating modes, complicating the device further. Assuming the modal shift in the central channel. We assume waveguide di-
nclad = 1.44398, Δin dex ∼ 0.25%, and dimensions of 5 mensions of 5 μm × 14 μm, and equally effective index spread
μm × 45 μm, the effective index spread is 0.8 × 10−3 . A between 1.5 and 1.6, (i.e., a huge total effective index spread
long period grating can serve as a mode converter, with the of 0.1) and fundamental diffraction order of 45. The compensa-
grating period satisfying the phase matching criterion. For our tion is perfect for the central channel frequency, however due to
effective index spread, average effective index and center oper- the diffraction order dependency of the spatial dispersions, non-
ating wavelength, we require a grating period of Λ = λc /Δn = central wavelengths will incur increasing modal offsets. Since
1931.25 μm. Assuming a mode coupling strenght of κ = 0.1 we enforced very high index contrast, the modal dependency
mm−1 [22], mode conversion can be supported over the C-band will be strong for non-central channels, limiting the efficiency
with 9 periods (total LPG length of 17.38 mm). With mid-way of this method to the center of the spectrum. For 37.5 GHz
mode conversion, the relatively high effective index spread can channel bandwidth only 5 channels out of 22 meet our MDL
be reduced by compensating the first and fourth modes, leav- and crosstalk requirements.
ing an effective index spread of only 2 × 10−4 between the
V. CONCLUSION
second and third modes (Fig. 8). An additional mode conver-
sion between these modes could copensate the modal spread The design of integrated wavelength multiplexer operating
completely. over MDM with spatial modes arranged along one transverse
axis normal to the dispersion axis, was presented and analyzed.
As standard AWGR rely on spatial separation between the spec-
E. Modal Offset Compensation Through Multiple Diffraction
tral channels, adding spatial modes result in a trade-off between
Orders spectral separation (i.e., channel count) and low MDL. In other
Our second solution for modal dispersion compensation is words, spectral dispersion along with modal dispersion leads to
using different diffraction orders per spatial modes. This way mode-dependent-spatial-shifts and as an outcome, to a reduction
neighboring diffraction orders can be used to counteract the in spectral resolution in comparison standard AWGR devices.
BLAU AND MAROM: WAVELENGTH DEMULTIPLEXER DESIGNS OPERATING OVER MULTIPLE SPATIAL MODES 4400111

Five different solutions were proposed and numerically sim- Thus we can derive the acquired phase for every frequency
ulated for support of four spatial modes in the telecom band. component ω and mode l (0, . . . , M-1):
Three methods focus on minimizing the effective index spread:   
Very low index contrast AWGR, selective excitation of lower dβ (l) 
(l)
Δϕ (ω, l) = β (ωc ) + (ω − ωc ) ΔL
modes within highly MM AWGR and weakly coupled multi core dω ω c
AWGR. Fabricating very low index contrast waveguides is chal-  (l)

(l)
2πnef f ng mλc
lenging due to the large dimensions of the waveguide, requiring = + (ω − ωc )
thick layer deposition (with good spatial homogeneity), and λc c (0)
nef f
 (l) 
deep etching (challenging within silica while maintaining side- (l) 
wall smoothness and verticality). The highly multimode wave-
nef f ng ω − ωc
= 2πm (0)
+ (0) (A2)
guide fabrication faces the challenge of depositing highly doped n n ωc
ef f ef f
silica to a height of 25 microns, however etching should be easier
The phase no longer depends solely on the spectral shift and
in this case. The fabrication of vertically stacked weakly coupled
group index; it further depends on the spatial mode’s propaga-
AWGR can be approached in two ways: alternate deposition of
tion properties. As an outcome, the spatial modes are dispersed
undoped and doped silica layers and a single deep etch process
at the output plane in addition to the desired spectral disper-
(challenging, as noted above) or etching each guiding layer af-
sion. In our case, the spatial mode displacement is undesired,
ter deposition, which introduces mask alignment challenges and
since we aim at spectral multiplexing only. Modal dispersion
likely requiring planarization steps after deposition layer follow-
at the output plane spreads the spectral channels, thus it limits
ing etching. The last two designs suggested compensation tech-
the spectral resolution, as well as cause non-uniform mode cou-
niques of the effective index spread by either mid-span mode
pling to the output waveguide. Hence, we wish to minimize the
conversion or the use of multiple AWGR diffraction orders. Mid-
modal dependency and maximize the spectral resolution. We
span mode conversion requires the inscription of long-period
will describe in detail the spatial and spectral shift at the AWGR
Bragg gratings [25], which could be applied to the waveguide
output plane immediately.
width definition (in lithography). Compensation via multiple
As the waveguide geometry is single-mode in the plane of
diffraction orders was found to be limited in spectral extent (i.e.,
dispersion, we are free of spatial-spectral coupling because of
total channel count), due to runoff of the different diffraction
the spatial mode structure when coupling to the output wave-
orders. Of the suggested techniques, we consider selective exci-
guide. Since the waveguides are single mode in the axis of spa-
tation of highly MM AWGR, weakly coupled multi core AWGR
tial dispersion, we conventionally approximate them to a spatial
and modal offset compensation through mid-span mode con-
Gaussian mode of width Δ, i.e., G(x/Δ). The Gaussian input is
version as the more promising solutions, which combine con-
Fourier transformed to a wide Gaussian by the first star coupler
ventional fabrication methods, standard fiber to chip coupling,
of radius R, then excites the waveguide array and accumulates
and low insertion and mode dependent loss mechanisms, though
the array’s phase delay Δϕ:
challenges still abound. However, all five designs have favorable


attributes and can lead the way to SDM integrated WDM demul- Δξnslab ξ − jp
Ψl (ξ, ω) ∝ G · G
triplexers on chip. A last potential solution is to utilize a free- λc R j
Δ
space based solution, which eliminates the wavevector/spatial
mode dependence (under the paraxial approximation). · exp (i · j · Δϕ) (A3)
where ξ is the Fourier plane spatial unit, p is the arrayed wave-
guide pitch, j is the waveguide identifier in the waveguide array
APPENDIX – ANALYSIS OF ARRAYED WAVEGUIDE GRATING and i is the imaginary unit. The second star coupler Fourier
ROUTER SUPPORTING MULTIPLE SPATIAL MODES transforms the signal above. The Fourier transform of multipli-
In our case of waveguides that are multi-moded, we need to cation of functions is the convolution of the Fourier transform
express the acquired phase delays per propagating mode and of each function. In addition we approximate the phase delay
frequency. Assuming the AWGR is designed for the fundamen- which is applied discretely to each waveguide, as a continuous,
tal mode, l = 0, the length difference between the waveguide linear function along ξ. Hence:
arms is:

Δξnslab
Ψl (x, ω) ∝ F.T. G ⊗
λc R
⎧ ⎫
Δϕ (ω, l = 0) = β (0) (ωc ) ΔL = 2πm ⎨

ξ


ξ − jp
⎬
2πm mλc F.T. exp i · · Δϕ · G
⇒ ΔL = = (0) (A1) ⎩ p Δ ⎭
j
β (0) (ωc ) nef f
  
Δξnslab ξ
= F.T. G ⊗ F.T. exp i · · Δϕ
λc R p
We denote the propagation constant and effective refractive ⎧ ⎫
(l)
index of the lth mode β (l) , nef f (e.g., β (0) is the propagation ⎨
ξ − jp ⎬
constant of the fundamental mode), and ωc and λc are the central ⊗ F.T. G (A4)
⎩ Δ ⎭
j
frequency/wavelength of the AWGR.
4400111 IEEE JOURNAL OF SELECTED TOPICS IN QUANTUM ELECTRONICS, VOL. 26, NO. 4, JULY/AUGUST 2020

The sum of Gaussians in the right expression, can be recast as Denoting the displacement of the diffraction orders by
a convolution of a Gaussian with a Dirak comb array function, xF S R = λc · R/(nslab · p):
or a sum of delta functions:  
Δ
   Ψl (x, ω) ∝ G x
Δ · ξnslab p · xF S R
Ψl (x, ω) ∝ F.T. G
λc R    (l) 
(l) 
   x xF S R nef f ng ω−ωc
ξ · G − j+m (0)
+ (0)
⊗ F.T. exp i · · Δϕ j
Δ Δ n n ωc
ef f ef f
p
⎛⎛ ⎞⎞ (A10)
 
ξ This expression consists of a wide Gaussian envelope, the
⊗ F.T. ⎝⎝G ⊗ δ (ξ − jp)⎠⎠
Δ j
Fourier transform of the individual waveguide output multi-
      plied by the dispersed signal, namely the sum of offset Gaus-
Δ · ξnslab ξ sians whose position is wavelength and mode dependent. The
∝ F.T. G ⊗ F.T. exp i · · Δϕ
λc R p (l) (0)
mode dependency is expressed by nef f /nef f , which results in
⎛ ⎛ ⎞⎞ a mode dependent translation, and the spectral dispersion by
  
ξ ∝ ng (ω − ωc )/ωc thus having unique dispersive strength. For
⊗ ⎝F.T. G · F.T. ⎝ δ (ξ − jp)⎠⎠ (A5)
Δ j
j = −m we get the main diffraction order, and sinceΔ ∼ p, the
first Gaussian is the window function suppressing the neighbor-
The Fourier transform of the two convolved functions is ing diffraction orders.
(l) (0)
the multiplication of their Fourier transforms. Using fx as the For nef f → nef f Eq. A10 converges to the description of an
Fourier pair variable to ξ (late we will convert fx to the real AWGR of the single mode case:
spatial coordinate x):
 
  Δ
λc R Ψl (x, ω) ∝ G x
Ψl (x, ω) ∝ G (Δ · fx ) · G fx p · xF S R
Δ · nslab
    x xF S R
 
ng

ω − ωc

Δϕ j · G − j+m 1+
⊗ δ fx − ⊗ δ fx − (A6) Δ Δ nef f ωc
p j
p j
(A11)
G(Δ · fx ) can be excluded from the sum over j since it is
slowly varying in space, and can be considered a constant value REFERENCES
over each waveguide. The convolution of a Gaussian function [1] D. M. Marom and M. Blau, “Switching solutions for WDM-SDM optical
with delta function simply translates the Gaussian to the delta networks,” IEEE Commun. Mag., vol. 53, no. 2 pp. 60–68, Feb. 2015.
[2] L. E. Nelson et al., “Spatial superchannel routing in a two-span ROADM
function’s center: system for space division multiplexing,” J. Lightw. Technol., vol. 32, no. 4,
pp. 783–789, Feb. 2014.
Ψl (x, ω) ∝ G (Δ · fx ) [3] M. D. Feuer et al., “Joint digital signal processing receivers for spa-
 λc R   tial superchannels,” IEEE Photon. Technol. Lett., vol. 24, no. 21,
Δϕ j pp. 1957–1960, Nov. 2012.
· G fx − − (A7) [4] D. M. Marom et al., “Wavelength-selective switch with direct few mode
j
Δ · nslab p p fiber integration,” Opt. Express, vol. 23, pp. 5723–5737, 2015.
[5] K. Igarashi et al., “Ultra-dense spatial-division-multiplexed optical
Converting back to the real spatial variable, x, using fiber transmission over 6-mode 19-core fibers,” Opt. Express, vol. 24,
pp. 10213–10231, 2016.
fx = nslab · x/(λc · R): [6] D. Dai, J. Wang, S. Chen, S. Wang, and S. He, “Monolithically inte-
  grated 64-channel silicon hybrid demultiplexer enabling simultaneous
Δ · nslab wavelength-and mode-division-multiplexing,” Laser Photon. Rev., vol. 9,
Ψl (x, ω) ∝ G x
λc R no. 3, pp. 339–344, 2015.
[7] D. Melati, A. Alippi, A. Annoni, N. Peserico, and A. Melloni, “In-
1  λc RΔϕ jλc R

tegrated all-optical MIMO demultiplexer for mode- and wavelength-
· G x− − (A8) division-multiplexed transmission,” Opt. Lett, vol. 42, pp. 342–345, 2017.
j
Δ nslab p nslab p [8] H. Chen et al., “Demonstration of a photonic integrated mode coupler
with 3.072 Tb/s MDM and WDM transmission over few-mode fiber,”
(l) (l)
presented at the 18th OptoElectron. Commun. Conf. Int. Conf. Photon.
ne f f
( ω −ω
ng Switching, 2013, Paper PD2_5.
Substituting Δϕ = 2πm( (0) + (0) ω c )), we obtain:
c
ne f f ne f f [9] G. Cincotti and T. Kodama, “Design of AWG devices for both frequency
and mode multiplexing,” in Proc. Extended Abstr. 24th Congr. Int. Com-
  
Δ · nslab x jλc R mun. Opt., Tokyo, 2017
Ψl (x, ω) ∝ G x · G − [10] H. Chen et al., “Multi-mode arrayed waveguide grating demultiplexer
λ0 R j
Δ nslab pΔ with single mode performance and few mode interfaces,” in Proc. Eur.
Conf. Opt. Commun., 2015, pp. 1–3.
 (l)   [11] A. A. M. Kok, S. Musa, A. Borreman, M. B. J. Diemeer, and
(l)
mλc R nef f ng ω − ωc A. Driessen, “Completely multimode arrayed waveguide grating-based
− (0)
+ (0)
(A9) wavelength demultiplexer,” in Proc. IEEE 8 EUROCON Comput. Tool,
nslab pΔ nef f nef f ωc
2003, pp. 422–426.
BLAU AND MAROM: WAVELENGTH DEMULTIPLEXER DESIGNS OPERATING OVER MULTIPLE SPATIAL MODES 4400111

[12] D. R. Drachenberg et al., “First selective mode excitation and amplification Miri Blau received the B.Sc. degree in electro optics engineering from the
in a ribbon core optical fiber,” Opt. Express, vol. 21, pp. 11257–11269, Jerusalem College of Technology, Jerusalem, Israel, in 2010. She is currently
2013. working toward the Ph.D. degree with the Department of Applied Physics,
[13] M. K. Smit, “New focusing and dispersive planar component based on Hebrew University, Jerusalem, Israel, under the supervision of Prof. Marom.
an optical phased array,” Electron. Lett., vol. 24, no. 7, pp. 385–386, Her research deals with space division multiplexing for overcoming the capacity
Mar. 1988. crunch accompanied by lower implementation costs than single-mode-based
[14] K. Okamoto, H. Takahashi, S. Suzuki, A. Sugita, and Y. Ohmori, “De- optical communications.
sign and fabrication of integrated-optic 8∗8 star coupler,” Electron. Lett.,
vol. 27, no. 9, pp. 774–775, Apr. 1991.
[15] C. Dragone, “Efficient N ∗ N star coupler based on Fourier optics,” Elec-
tron. Lett., vol. 24, no. 15, pp. 942–944, 1988.
[16] P. Munoz, D. Pastor, and J. Capmany, “Modeling and design of arrayed
waveguide gratings,” J. Lightw. Technol., vol. 20, no. 4, pp. 661–674,
Apr. 2002.
[17] H. Takahashi, Y. Hibino, and I. Nishi, “Polarization-insensitive arrayed-
waveguide grating wavelength multiplexer on silicon,” Opt. Lett., vol. 17,
pp. 499–501, 1992. Dan M. Marom (S’99–M’00–SM’08) received the B.Sc. degree in mechanical
[18] S. Ö. Arık, D. Askarov, and J. M. Kahn, “Effect of mode coupling on engineering and the M.Sc. degree in electrical engineering, both from Tel-Aviv
signal processing complexity in mode-division multiplexing,” J. Lightw. University, Tel Aviv, Israel, in 1989 and 1995, respectively, and received the
Technol, vol. 31, no. 3, pp. 423–431, Feb. 2013. Ph.D. degree in electrical engineering from the University of California, San
[19] D. Ohana, B. Desiatov, N. Mazurski, and U. Levy, “Dielectric metasurface Diego (UCSD), San Diego, CA, USA, in 2000. From 2000 until 2005, he was a
as a platform for spatial mode conversion in nanoscale waveguides,” Nano Member of the Technical Staff with the Advanced Photonics Research Depart-
Lett., vol. 16, no. 12, pp. 7956–7961, 2016. ment, Bell Laboratories, Lucent Technologies, where he invented and headed the
[20] L. Rechtman and D. M. Marom, “Rectangular versus circular fiber core research and development efforts of MEMS-based wavelength-selective switch-
designs: New opportunities for mode division multiplexing?,” in Proc. ing solutions for optical networks. Since 2005, he has been with the Applied
Opt. Fiber Commun. Conf. Exhib., 2017, pp. 1–3. Physics Department, The Hebrew University of Jerusalem, Israel, where he is
[21] K. Hammani et al., “Towards nonlinear conversion from mid- to currently serving as Department Chair and leads the Photonics Devices group.
near-infrared wavelengths using Silicon Germanium waveguides,” Opt. His research interests include creating photonic devices and sub-systems for
Express, vol. 22, pp. 9667–9674, 2014. switching and manipulating optical signals in guided-wave and free-space op-
[22] G. L. Bona, R. Germann, and B. J. Offrein, “SiON high-refractive-index tics solutions using light modulating devices, nonlinear optics, and compound
waveguide and planar lightwave circuits,” IBM J. Res. Develop., vol. 47, materials.
no. 2.3, pp. 239–249, Mar. 2003. Prof. Marom is a Fellow of the Optical Society of America, a Senior Member
[23] D. Duchesne et al., “Efficient self-phase modulation in low loss, high of the IEEE Photonics Society and an Elected Member of that Society’s Board
index doped silica glass integrated waveguides,” Opt. Express, vol. 17, of Governors. He was the recipient of the IEEE Photonics Society Distinguished
pp. 1865–1870, 2009. Lecturer Award in 2014 and 2015. From 1996 to 2000, he was a Fannie and
[24] S. H. Nam et al., “Bend-insensitive ultra short long-period gratings by the John Hertz Foundation Graduate Fellow with UCSD, and was a Peter Brojde
electric arc method and their applications to harsh environment sensing Scholar in 2006–2007. He was one of two Senior Editors for Photonics Tech-
and communication,” Opt. Express, vol. 13, pp. 731–737, 2005. nology Letters, handling all photonic-device-related submissions. He serves as
[25] D. Ohana and U. Levy, “Mode conversion based on dielectric metamaterial Steering Committee Chair for the International Conference on Optical MEMS
in silicon,” Opt. Express, vol. 22, pp. 27617–27631, 2014. and Nanophotonics.

Potrebbero piacerti anche